Validation of a multimodal AI biomarker in a Puerto Rican cohort with localized prostate cancer

Introduction/Background: Puerto Rican men face the highest prostate cancer mortality among Hispanic subgroups, exceeding rates in non-Hispanic Black populations. Despite this disproportionate burden, validated prognostic tools generalized for this demographic are currently absent from literature. We addressed this by evaluating a multimodal artificial intelligence (MMAI) biomarker in a representative Puerto Rican real-world cohort.

Methods/Design: This retrospective, multi-site study analyzed 284 Puerto Rican patients with localized prostate cancer (2012-2019). The MMAI biomarker (v1.2) integrated digitized biopsy images with baseline variables (age, PSA, and T-stage). Associations with distant metastasis (DM) and recurrence (BCR) were evaluated continuously (per 1 SD) and categorically (low, intermediate, high) using univariable/multivariable Cox models and Kaplan-Meier estimates.

Results: The cohort spanned NCCN low- (38%), intermediate- (42%), and high-risk (20%) categories. Treatments included surgery (42%), radiation (54%), and non-curative management (AS or ADT monotherapy) (5%). With a 6-year median follow-up, MMAI scores significantly predicted DM (HR 1.74; 95% CI: 1.05-2.88; p=0.03) and BCR (HR 1.66; 95% CI: 1.29-2.14; p<0.001). Ten-year freedom from BCR for MMAI low-, intermediate-, and high-risk groups was 91% (86-95%), 66% (47-80%), and 34% (7-64%), respectively. Ten-year freedom from DM was 97% (89-99%), 90% (74-97%), and 87% (52-97%). MMAI provided significant risk reclassification within NCCN categories (p<0.001).

Conclusion/Discussion: This study provides the first validation of a MMAI biomarker in a Hispanic Puerto Rican cohort, demonstrating independent prognostic performance for DM and BCR. These results broaden the generalizability of the MMAI model and support its ability to personalize care in this underserved, high-risk population.

Learning Objectives:

1. Understand the need for population-specific AI tools in Puerto Rican prostate cancer outcomes.
2. Evaluate MMAI biomarker performance in predicting metastasis and recurrence risk.
3. Assess how MMAI improves risk stratification beyond NCCN categories for personalized care.